Carbon emissions from AI in 2025 will equal New York’s level
A study showed that artificial intelligence (AI) in 2025 could produce carbon dioxide emissions at the level of 52.2 million tons, which is equivalent to the emissions of New York. At the same time, water consumption for its cooling could exceed the figures of the entire global bottled water industry.
The research, conducted by Dutch scientist Alex de Vries-Gao, evaluates the side effects of digital technologies. According to his findings, the carbon footprint of AI in 2025 could increase to 80 million tons of CO2, which is equivalent to more than 8% of global aviation emissions. Meanwhile, in Ukraine, where greenhouse gas emissions have also increased due to military actions and forest fires, a similar situation was observed.
Additionally, the study for the first time assessed the water footprint of AI. It is projected that AI systems in 2025 could consume up to 765 billion liters of water needed to cool the data centers that power them. This estimate exceeds by more than a third the previous estimates of water consumption by all data centers combined.
According to data from the International Energy Agency (IEA), AI-oriented databases consume as much electricity as large industrial enterprises, and by 2030 this consumption could double. The largest contributors to global resource consumption are the USA (45%), China (25%), and Europe (15%).
In light of this, in November 2025 Reuters reported a malfunction in the largest exchange operator CME Group due to a cooling failure in one of the data centers, highlighting the vulnerability of the existing infrastructure.
| Country | Share in AI energy consumption, % |
| USA | 45 |
| China | 25 |
| Europe | 15 |
